Show More
@@ -1,367 +1,368 b'' | |||
|
1 | 1 | # ancestor.py - generic DAG ancestor algorithm for mercurial |
|
2 | 2 | # |
|
3 | 3 | # Copyright 2006 Matt Mackall <mpm@selenic.com> |
|
4 | 4 | # |
|
5 | 5 | # This software may be used and distributed according to the terms of the |
|
6 | 6 | # GNU General Public License version 2 or any later version. |
|
7 | 7 | |
|
8 | 8 | from __future__ import absolute_import |
|
9 | 9 | |
|
10 | 10 | import heapq |
|
11 | 11 | |
|
12 | 12 | from .node import nullrev |
|
13 | 13 | from . import ( |
|
14 | 14 | pycompat, |
|
15 | 15 | ) |
|
16 | 16 | |
|
17 | 17 | def commonancestorsheads(pfunc, *nodes): |
|
18 | 18 | """Returns a set with the heads of all common ancestors of all nodes, |
|
19 | 19 | heads(::nodes[0] and ::nodes[1] and ...) . |
|
20 | 20 | |
|
21 | 21 | pfunc must return a list of parent vertices for a given vertex. |
|
22 | 22 | """ |
|
23 | 23 | if not isinstance(nodes, set): |
|
24 | 24 | nodes = set(nodes) |
|
25 | 25 | if nullrev in nodes: |
|
26 | 26 | return set() |
|
27 | 27 | if len(nodes) <= 1: |
|
28 | 28 | return nodes |
|
29 | 29 | |
|
30 | 30 | allseen = (1 << len(nodes)) - 1 |
|
31 | 31 | seen = [0] * (max(nodes) + 1) |
|
32 | 32 | for i, n in enumerate(nodes): |
|
33 | 33 | seen[n] = 1 << i |
|
34 | 34 | poison = 1 << (i + 1) |
|
35 | 35 | |
|
36 | 36 | gca = set() |
|
37 | 37 | interesting = len(nodes) |
|
38 | 38 | nv = len(seen) - 1 |
|
39 | 39 | while nv >= 0 and interesting: |
|
40 | 40 | v = nv |
|
41 | 41 | nv -= 1 |
|
42 | 42 | if not seen[v]: |
|
43 | 43 | continue |
|
44 | 44 | sv = seen[v] |
|
45 | 45 | if sv < poison: |
|
46 | 46 | interesting -= 1 |
|
47 | 47 | if sv == allseen: |
|
48 | 48 | gca.add(v) |
|
49 | 49 | sv |= poison |
|
50 | 50 | if v in nodes: |
|
51 | 51 | # history is linear |
|
52 | 52 | return {v} |
|
53 | 53 | if sv < poison: |
|
54 | 54 | for p in pfunc(v): |
|
55 | 55 | sp = seen[p] |
|
56 | 56 | if p == nullrev: |
|
57 | 57 | continue |
|
58 | 58 | if sp == 0: |
|
59 | 59 | seen[p] = sv |
|
60 | 60 | interesting += 1 |
|
61 | 61 | elif sp != sv: |
|
62 | 62 | seen[p] |= sv |
|
63 | 63 | else: |
|
64 | 64 | for p in pfunc(v): |
|
65 | 65 | if p == nullrev: |
|
66 | 66 | continue |
|
67 | 67 | sp = seen[p] |
|
68 | 68 | if sp and sp < poison: |
|
69 | 69 | interesting -= 1 |
|
70 | 70 | seen[p] = sv |
|
71 | 71 | return gca |
|
72 | 72 | |
|
73 | 73 | def ancestors(pfunc, *orignodes): |
|
74 | 74 | """ |
|
75 | 75 | Returns the common ancestors of a and b that are furthest from a |
|
76 | 76 | root (as measured by longest path). |
|
77 | 77 | |
|
78 | 78 | pfunc must return a list of parent vertices for a given vertex. |
|
79 | 79 | """ |
|
80 | 80 | def deepest(nodes): |
|
81 | 81 | interesting = {} |
|
82 | 82 | count = max(nodes) + 1 |
|
83 | 83 | depth = [0] * count |
|
84 | 84 | seen = [0] * count |
|
85 | 85 | mapping = [] |
|
86 | 86 | for (i, n) in enumerate(sorted(nodes)): |
|
87 | 87 | depth[n] = 1 |
|
88 | 88 | b = 1 << i |
|
89 | 89 | seen[n] = b |
|
90 | 90 | interesting[b] = 1 |
|
91 | 91 | mapping.append((b, n)) |
|
92 | 92 | nv = count - 1 |
|
93 | 93 | while nv >= 0 and len(interesting) > 1: |
|
94 | 94 | v = nv |
|
95 | 95 | nv -= 1 |
|
96 | 96 | dv = depth[v] |
|
97 | 97 | if dv == 0: |
|
98 | 98 | continue |
|
99 | 99 | sv = seen[v] |
|
100 | 100 | for p in pfunc(v): |
|
101 | 101 | if p == nullrev: |
|
102 | 102 | continue |
|
103 | 103 | dp = depth[p] |
|
104 | 104 | nsp = sp = seen[p] |
|
105 | 105 | if dp <= dv: |
|
106 | 106 | depth[p] = dv + 1 |
|
107 | 107 | if sp != sv: |
|
108 | 108 | interesting[sv] += 1 |
|
109 | 109 | nsp = seen[p] = sv |
|
110 | 110 | if sp: |
|
111 | 111 | interesting[sp] -= 1 |
|
112 | 112 | if interesting[sp] == 0: |
|
113 | 113 | del interesting[sp] |
|
114 | 114 | elif dv == dp - 1: |
|
115 | 115 | nsp = sp | sv |
|
116 | 116 | if nsp == sp: |
|
117 | 117 | continue |
|
118 | 118 | seen[p] = nsp |
|
119 | 119 | interesting.setdefault(nsp, 0) |
|
120 | 120 | interesting[nsp] += 1 |
|
121 | 121 | interesting[sp] -= 1 |
|
122 | 122 | if interesting[sp] == 0: |
|
123 | 123 | del interesting[sp] |
|
124 | 124 | interesting[sv] -= 1 |
|
125 | 125 | if interesting[sv] == 0: |
|
126 | 126 | del interesting[sv] |
|
127 | 127 | |
|
128 | 128 | if len(interesting) != 1: |
|
129 | 129 | return [] |
|
130 | 130 | |
|
131 | 131 | k = 0 |
|
132 | 132 | for i in interesting: |
|
133 | 133 | k |= i |
|
134 | 134 | return set(n for (i, n) in mapping if k & i) |
|
135 | 135 | |
|
136 | 136 | gca = commonancestorsheads(pfunc, *orignodes) |
|
137 | 137 | |
|
138 | 138 | if len(gca) <= 1: |
|
139 | 139 | return gca |
|
140 | 140 | return deepest(gca) |
|
141 | 141 | |
|
142 | 142 | class incrementalmissingancestors(object): |
|
143 | 143 | '''persistent state used to calculate missing ancestors incrementally |
|
144 | 144 | |
|
145 | 145 | Although similar in spirit to lazyancestors below, this is a separate class |
|
146 | 146 | because trying to support contains and missingancestors operations with the |
|
147 | 147 | same internal data structures adds needless complexity.''' |
|
148 | 148 | def __init__(self, pfunc, bases): |
|
149 | 149 | self.bases = set(bases) |
|
150 | 150 | if not self.bases: |
|
151 | 151 | self.bases.add(nullrev) |
|
152 | 152 | self.pfunc = pfunc |
|
153 | 153 | |
|
154 | 154 | def hasbases(self): |
|
155 | 155 | '''whether the common set has any non-trivial bases''' |
|
156 | 156 | return self.bases and self.bases != {nullrev} |
|
157 | 157 | |
|
158 | 158 | def addbases(self, newbases): |
|
159 | 159 | '''grow the ancestor set by adding new bases''' |
|
160 | 160 | self.bases.update(newbases) |
|
161 | 161 | |
|
162 | 162 | def removeancestorsfrom(self, revs): |
|
163 | 163 | '''remove all ancestors of bases from the set revs (in place)''' |
|
164 | 164 | bases = self.bases |
|
165 | 165 | pfunc = self.pfunc |
|
166 | 166 | revs.difference_update(bases) |
|
167 | 167 | # nullrev is always an ancestor |
|
168 | 168 | revs.discard(nullrev) |
|
169 | 169 | if not revs: |
|
170 | 170 | return |
|
171 | 171 | # anything in revs > start is definitely not an ancestor of bases |
|
172 | 172 | # revs <= start needs to be investigated |
|
173 | 173 | start = max(bases) |
|
174 | 174 | keepcount = sum(1 for r in revs if r > start) |
|
175 | 175 | if len(revs) == keepcount: |
|
176 | 176 | # no revs to consider |
|
177 | 177 | return |
|
178 | 178 | |
|
179 | 179 | for curr in pycompat.xrange(start, min(revs) - 1, -1): |
|
180 | 180 | if curr not in bases: |
|
181 | 181 | continue |
|
182 | 182 | revs.discard(curr) |
|
183 | 183 | bases.update(pfunc(curr)) |
|
184 | 184 | if len(revs) == keepcount: |
|
185 | 185 | # no more potential revs to discard |
|
186 | 186 | break |
|
187 | 187 | |
|
188 | 188 | def missingancestors(self, revs): |
|
189 | 189 | '''return all the ancestors of revs that are not ancestors of self.bases |
|
190 | 190 | |
|
191 | 191 | This may include elements from revs. |
|
192 | 192 | |
|
193 | 193 | Equivalent to the revset (::revs - ::self.bases). Revs are returned in |
|
194 | 194 | revision number order, which is a topological order.''' |
|
195 | 195 | revsvisit = set(revs) |
|
196 | 196 | basesvisit = self.bases |
|
197 | 197 | pfunc = self.pfunc |
|
198 | 198 | bothvisit = revsvisit.intersection(basesvisit) |
|
199 | 199 | revsvisit.difference_update(bothvisit) |
|
200 | 200 | if not revsvisit: |
|
201 | 201 | return [] |
|
202 | 202 | |
|
203 | 203 | start = max(max(revsvisit), max(basesvisit)) |
|
204 | 204 | # At this point, we hold the invariants that: |
|
205 | 205 | # - revsvisit is the set of nodes we know are an ancestor of at least |
|
206 | 206 | # one of the nodes in revs |
|
207 | 207 | # - basesvisit is the same for bases |
|
208 | 208 | # - bothvisit is the set of nodes we know are ancestors of at least one |
|
209 | 209 | # of the nodes in revs and one of the nodes in bases. bothvisit and |
|
210 | 210 | # revsvisit are mutually exclusive, but bothvisit is a subset of |
|
211 | 211 | # basesvisit. |
|
212 | 212 | # Now we walk down in reverse topo order, adding parents of nodes |
|
213 | 213 | # already visited to the sets while maintaining the invariants. When a |
|
214 | 214 | # node is found in both revsvisit and basesvisit, it is removed from |
|
215 | 215 | # revsvisit and added to bothvisit. When revsvisit becomes empty, there |
|
216 | 216 | # are no more ancestors of revs that aren't also ancestors of bases, so |
|
217 | 217 | # exit. |
|
218 | 218 | |
|
219 | 219 | missing = [] |
|
220 | 220 | for curr in pycompat.xrange(start, nullrev, -1): |
|
221 | 221 | if not revsvisit: |
|
222 | 222 | break |
|
223 | 223 | |
|
224 | 224 | if curr in bothvisit: |
|
225 | 225 | bothvisit.remove(curr) |
|
226 | 226 | # curr's parents might have made it into revsvisit through |
|
227 | 227 | # another path |
|
228 | 228 | for p in pfunc(curr): |
|
229 | 229 | revsvisit.discard(p) |
|
230 | 230 | basesvisit.add(p) |
|
231 | 231 | bothvisit.add(p) |
|
232 | 232 | continue |
|
233 | 233 | |
|
234 | 234 | if curr in revsvisit: |
|
235 | 235 | missing.append(curr) |
|
236 | 236 | revsvisit.remove(curr) |
|
237 | 237 | thisvisit = revsvisit |
|
238 | 238 | othervisit = basesvisit |
|
239 | 239 | elif curr in basesvisit: |
|
240 | 240 | thisvisit = basesvisit |
|
241 | 241 | othervisit = revsvisit |
|
242 | 242 | else: |
|
243 | 243 | # not an ancestor of revs or bases: ignore |
|
244 | 244 | continue |
|
245 | 245 | |
|
246 | 246 | for p in pfunc(curr): |
|
247 | 247 | if p == nullrev: |
|
248 | 248 | pass |
|
249 | 249 | elif p in othervisit or p in bothvisit: |
|
250 | 250 | # p is implicitly in thisvisit. This means p is or should be |
|
251 | 251 | # in bothvisit |
|
252 | 252 | revsvisit.discard(p) |
|
253 | 253 | basesvisit.add(p) |
|
254 | 254 | bothvisit.add(p) |
|
255 | 255 | else: |
|
256 | 256 | # visit later |
|
257 | 257 | thisvisit.add(p) |
|
258 | 258 | |
|
259 | 259 | missing.reverse() |
|
260 | 260 | return missing |
|
261 | 261 | |
|
262 | 262 | # Extracted from lazyancestors.__iter__ to avoid a reference cycle |
|
263 | 263 | def _lazyancestorsiter(parentrevs, initrevs, stoprev, inclusive): |
|
264 | 264 | seen = {nullrev} |
|
265 | 265 | schedule = heapq.heappush |
|
266 | 266 | nextitem = heapq.heappop |
|
267 | 267 | see = seen.add |
|
268 | 268 | |
|
269 | 269 | if inclusive: |
|
270 | 270 | visit = [-r for r in initrevs] |
|
271 | 271 | seen.update(initrevs) |
|
272 | 272 | heapq.heapify(visit) |
|
273 | 273 | else: |
|
274 | 274 | visit = [] |
|
275 | 275 | heapq.heapify(visit) |
|
276 | 276 | for r in initrevs: |
|
277 | 277 | for parent in parentrevs(r): |
|
278 | 278 | if parent not in seen: |
|
279 | 279 | schedule(visit, -parent) |
|
280 | 280 | see(parent) |
|
281 | 281 | |
|
282 | 282 | while visit: |
|
283 | 283 | current = -nextitem(visit) |
|
284 |
if current |
|
|
285 |
|
|
|
286 | for parent in parentrevs(current): | |
|
287 | if parent not in seen: | |
|
288 | schedule(visit, -parent) | |
|
289 |
|
|
|
284 | if current < stoprev: | |
|
285 | break | |
|
286 | yield current | |
|
287 | for parent in parentrevs(current): | |
|
288 | if parent not in seen: | |
|
289 | schedule(visit, -parent) | |
|
290 | see(parent) | |
|
290 | 291 | |
|
291 | 292 | class lazyancestors(object): |
|
292 | 293 | def __init__(self, pfunc, revs, stoprev=0, inclusive=False): |
|
293 | 294 | """Create a new object generating ancestors for the given revs. Does |
|
294 | 295 | not generate revs lower than stoprev. |
|
295 | 296 | |
|
296 | 297 | This is computed lazily starting from revs. The object supports |
|
297 | 298 | iteration and membership. |
|
298 | 299 | |
|
299 | 300 | cl should be a changelog and revs should be an iterable. inclusive is |
|
300 | 301 | a boolean that indicates whether revs should be included. Revs lower |
|
301 | 302 | than stoprev will not be generated. |
|
302 | 303 | |
|
303 | 304 | Result does not include the null revision.""" |
|
304 | 305 | self._parentrevs = pfunc |
|
305 | 306 | self._initrevs = revs = [r for r in revs if r >= stoprev] |
|
306 | 307 | self._stoprev = stoprev |
|
307 | 308 | self._inclusive = inclusive |
|
308 | 309 | |
|
309 | 310 | self._containsseen = set() |
|
310 | 311 | self._containsiter = _lazyancestorsiter(self._parentrevs, |
|
311 | 312 | self._initrevs, |
|
312 | 313 | self._stoprev, |
|
313 | 314 | self._inclusive) |
|
314 | 315 | |
|
315 | 316 | def __nonzero__(self): |
|
316 | 317 | """False if the set is empty, True otherwise.""" |
|
317 | 318 | try: |
|
318 | 319 | next(iter(self)) |
|
319 | 320 | return True |
|
320 | 321 | except StopIteration: |
|
321 | 322 | return False |
|
322 | 323 | |
|
323 | 324 | __bool__ = __nonzero__ |
|
324 | 325 | |
|
325 | 326 | def __iter__(self): |
|
326 | 327 | """Generate the ancestors of _initrevs in reverse topological order. |
|
327 | 328 | |
|
328 | 329 | If inclusive is False, yield a sequence of revision numbers starting |
|
329 | 330 | with the parents of each revision in revs, i.e., each revision is |
|
330 | 331 | *not* considered an ancestor of itself. Results are emitted in reverse |
|
331 | 332 | revision number order. That order is also topological: a child is |
|
332 | 333 | always emitted before its parent. |
|
333 | 334 | |
|
334 | 335 | If inclusive is True, the source revisions are also yielded. The |
|
335 | 336 | reverse revision number order is still enforced.""" |
|
336 | 337 | for rev in _lazyancestorsiter(self._parentrevs, self._initrevs, |
|
337 | 338 | self._stoprev, self._inclusive): |
|
338 | 339 | yield rev |
|
339 | 340 | |
|
340 | 341 | def __contains__(self, target): |
|
341 | 342 | """Test whether target is an ancestor of self._initrevs.""" |
|
342 | 343 | seen = self._containsseen |
|
343 | 344 | if target in seen: |
|
344 | 345 | return True |
|
345 | 346 | iter = self._containsiter |
|
346 | 347 | if iter is None: |
|
347 | 348 | # Iterator exhausted |
|
348 | 349 | return False |
|
349 | 350 | # Only integer target is valid, but some callers expect 'None in self' |
|
350 | 351 | # to be False. So we explicitly allow it. |
|
351 | 352 | if target is None: |
|
352 | 353 | return False |
|
353 | 354 | |
|
354 | 355 | see = seen.add |
|
355 | 356 | try: |
|
356 | 357 | while True: |
|
357 | 358 | rev = next(iter) |
|
358 | 359 | see(rev) |
|
359 | 360 | if rev == target: |
|
360 | 361 | return True |
|
361 | 362 | if rev < target: |
|
362 | 363 | return False |
|
363 | 364 | except StopIteration: |
|
364 | 365 | # Set to None to indicate fast-path can be used next time, and to |
|
365 | 366 | # free up memory. |
|
366 | 367 | self._containsiter = None |
|
367 | 368 | return False |
General Comments 0
You need to be logged in to leave comments.
Login now